Pavel Roskin

Results of `./mfakto -st2`: ``` Self-test statistics number of tests 335250 successful tests 303138 no factor found 32112 self-test FAILED! ERROR: self-test failed, exiting. ``` I see some failures from...

Confirmed that all errors are from cl_barrett15_74_gs. It's interesting that the number of "no factor found" errors is 32100 now, 12 less than the previous time. There were 78 successes...

Closing in favor of https://github.com/primesearch/mfakto/pull/16

It there any test case that would fail without the barrett15 fix (with double precision) and pass with the fix? It should be added to catch similar issues in the...

> The latest changes by @brubsby should fix the self-test failures. I consistently had four self-tests fail on Intel integrated GPUs on macOS, and Tyler's fork does have this issue....

The errors look like what I'm getting with `VectorSize=2`, but only for `cl_barrett32_87_gs`. Do you mind to try `VectorSize=1`? Speaking of strange things, I believe mfakto was working on the...

Yes, I can trigger the warning by setting `GPUType=GCN`: ``` WARNING: VectorSize=1 is known to fail on AMD GPUs and drivers. If the selftest fails, please increase VectorSize to 2...

I was able to "fix" the issue by commenting out all failing kernels for the `INTEL` GPUs. The self-test started passing once I commented out every `MUL32`: ```diff diff --git...

I tried Mesa instead of the Intel Compute Runtime, and the failures are gone! That seems to imply that they are in software that runs on the CPU, not in...

The issue with 15-bit barrett kernels and VectorSize=1 is fixed in #46. It's unrelated to the hardware, I can reproduce it in pocl. The issue with 32-bit barrett kernels is...